feat(ui): solo-find bonus — reward a model for catching what others missed

Adds an editable 'solo-find bonus ×' (default 1.5). A confirmed finding reported by exactly one model (derived from the global reporter count per content-addressed finding — no grader flagging needed) scores severity × bonus. New 'solo' column counts uniquely-caught confirmed findings. Solo-ness is computed over ALL data so the model filter can't fake it. Client-side only; store stays point-free.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

README.md

@@ -137,6 +137,12 @@ Blocking→high, Minor→small): `penalty × points[claimed]`. So a Blocking-cla
 `high(8) × -0.5 = -4`, and a model with the odd good find but many false positives nets *down* —
 even negative — instead of coasting on its hits.
 
+And an editable **solo-find bonus ×** (default `1.5`). Because findings are content-addressed, the
+number of models that reported one is known, so a confirmed finding that **only that model** caught
+(no other model reported it) scores `severity × bonus` — rewarding catching what the swarm missed.
+The `solo` column counts those. This is derived from the data (reporter count); the grader never has
+to flag it. Set the bonus to `1` to disable.
+
 Auth: the `/ui` shell is public (it holds no data); paste the store token into its **connect** box,
 or open `/ui?token=<token>` once (remembered in `localStorage`). Prefer your own dashboard? Point
 Grafana/Metabase/etc. at the SQLite file or the same `/export` + `/scoreboard` + `/runs` JSON.

ui.html

@@ -80,6 +80,7 @@
         <span class="small mut">high</span><input type="number" id="p_high" value="8">
         <span class="small mut">critical</span><input type="number" id="p_critical" value="20">
         <span class="small mut" style="margin-left:18px">false-positive penalty ×</span><input type="number" id="fp_mult" value="-0.5" step="0.5" title="A false positive scores this × the severity the model CLAIMED (its lens verdict). e.g. a Blocking-claimed FP at -0.5 = high(8) × -0.5 = -4 pts.">
+        <span class="small mut" style="margin-left:18px">solo-find bonus ×</span><input type="number" id="solo_bonus" value="1.5" step="0.5" min="1" title="A confirmed finding that NO other model reported scores this × its severity points — rewarding a model for catching what the swarm missed. 1 = no bonus.">
       </div>
     </div>
   </div>
@@ -166,6 +167,7 @@ function curve(){
   return c;
 }
 function fpMult(){ const v = parseFloat(document.getElementById("fp_mult").value); return isNaN(v) ? 0 : v; }
+function soloBonus(){ const v = parseFloat(document.getElementById("solo_bonus").value); return isNaN(v) ? 1 : v; }
 // A false positive has no graded severity, so penalize it by the severity the
 // MODEL claimed — its lens verdict (raw_severity) — mapped onto the curve. The
 // louder the wrong cry, the bigger the penalty.
@@ -214,9 +216,13 @@ function rowMatch(row, f){
 
 function aggregate(f){
   const c = curve();
+  // GLOBAL reporter set per finding (ignores filters) — a finding is "solo" when
+  // exactly one model ever reported it, so the model filter can't fake solo-ness.
+  const reporters = new Map();
+  for (const r of ROWS){ if(!reporters.has(r.finding_id)) reporters.set(r.finding_id, new Set()); reporters.get(r.finding_id).add(r.model); }
   const M = new Map();
   const get = m => { if(!M.has(m)) M.set(m, {model:m, provider:"", runs:0, minutes:0, inTok:0, outTok:0,
-    findings:new Set(), confirmed:new Set(), fp:new Map(), ungraded:new Set(), sev:Object.fromEntries(SEVS.map(s=>[s,new Set()]))}); return M.get(m); };
+    findings:new Set(), confirmed:new Map(), fp:new Map(), ungraded:new Set()}); return M.get(m); };
 
   for (const r of RUNS){ if(!runMatch(r,f)) continue; const m=get(r.model); m.runs++; m.minutes += (r.duration_secs||0)/60;
     m.inTok += r.input_tokens||0; m.outTok += r.output_tokens||0; if(r.provider) m.provider=r.provider; }
@@ -224,20 +230,26 @@ function aggregate(f){
   const rows = ROWS.filter(r => rowMatch(r, f));
   for (const r of rows){ const m=get(r.model); if(r.provider) m.provider=m.provider||r.provider;
     m.findings.add(r.finding_id);
-    if (r.graded && r.is_real === true){ m.confirmed.add(r.finding_id); if (r.severity) m.sev[r.severity].add(r.finding_id); }
+    if (r.graded && r.is_real === true){ m.confirmed.set(r.finding_id, r.severity || ""); }
     else if (r.graded && r.is_real === false){ m.fp.set(r.finding_id, rawToSevKey(r.raw_severity)); }
     else { m.ungraded.add(r.finding_id); }
   }
 
-  const fpm = fpMult();
+  const fpm = fpMult(), sb = soloBonus();
   const out = [...M.values()].map(m => {
-    const sevCounts = Object.fromEntries(SEVS.map(s=>[s, m.sev[s].size]));
-    const confirmedPoints = SEVS.reduce((a,s)=> a + c[s]*sevCounts[s], 0);
+    const sevCounts = Object.fromEntries(SEVS.map(s=>[s,0]));
+    let confirmedPoints = 0, solo = 0;
+    for (const [fid, sevv] of m.confirmed){
+      if (sevCounts[sevv] !== undefined) sevCounts[sevv]++;
+      const isSolo = (reporters.get(fid)?.size || 1) === 1; // only this model ever reported it
+      if (isSolo) solo++;
+      confirmedPoints += (c[sevv] || 0) * (isSolo ? sb : 1);
+    }
     let fpPen = 0; for (const k of m.fp.values()) fpPen += (c[k]||0) * fpm; // negative when fpm<0
-    const points = confirmedPoints + fpPen; // NET of the false-positive penalty
+    const points = confirmedPoints + fpPen; // NET: solo-boosted confirmed + FP penalty
     const findings = m.findings.size, confirmed = m.confirmed.size;
     return { model:m.model, provider:m.provider, runs:m.runs, minutes:m.minutes,
-      inTok:m.inTok, outTok:m.outTok, findings, confirmed, fp:m.fp.size, ungraded:m.ungraded.size,
+      inTok:m.inTok, outTok:m.outTok, findings, confirmed, solo, fp:m.fp.size, ungraded:m.ungraded.size,
       sev:sevCounts, confirmedPoints, fpPen, points,
       ptsPerMin: m.minutes>0 ? points/m.minutes : null,
       ptsPerRun: m.runs>0 ? points/m.runs : null,
@@ -249,7 +261,7 @@ function aggregate(f){
 const COLS = [
   {k:"model", t:"model", l:true}, {k:"provider", t:"provider", l:true},
   {k:"runs", t:"runs"}, {k:"minutes", t:"min", fmt:v=>v.toFixed(1)},
-  {k:"findings", t:"findings"}, {k:"confirmed", t:"real"}, {k:"fp", t:"FP"}, {k:"ungraded", t:"ungr"},
+  {k:"findings", t:"findings"}, {k:"confirmed", t:"real"}, {k:"solo", t:"solo"}, {k:"fp", t:"FP"}, {k:"ungraded", t:"ungr"},
   {k:"confirmedPct", t:"real%", fmt:v=>v==null?"—":v.toFixed(0)+"%"},
   {k:"fpPen", t:"fp pen", fmt:v=>v?v.toFixed(1):"0"},
   {k:"points", t:"points (net)", fmt:v=>v.toFixed(0)},
@@ -286,6 +298,7 @@ function render(){
       td.innerHTML = col.fmt ? col.fmt(v) : (v==null?"—":v);
       if ((col.k==="ptsPerMin" || col.k==="ptsPerRun" || col.k==="points") && v!=null) td.classList.add(v<0 ? "bad" : "good");
       if (col.k==="fpPen" && v<0) td.classList.add("bad");
+      if (col.k==="solo" && v>0) td.classList.add("good");
       if (col.k==="fp" && v>0) td.classList.add("warn");
       tr.appendChild(td);
     }